Scott Alexander

And for adults, we have numeracy and literacy, which have also gone down, with literacy going down sharply after 2014.

Cremier thinks this might be fake.

Link in post.

He says part of the effect is demographic shift.

Blacks, Hispanics and some Middle Eastern populations tend to underperform whites on most scholastic tests.

If they're recent immigrants, they might not even speak the language fluently.

As these groups increase in proportion of the test-taking population, test scores go down.

There's also a more arcane issue called measurement invariance.

Click the link for the explanation.

Cremier finds that when you adjust for these things, some of the problem goes away.

Here's a chart by Cremier.

How American PISA performance is trending.

Linked and demographically matched scores for 2000 to 2022.

So the y-axis is PISA subtest scores, and the x-axis is years.

For mathematics, we see the raw score trend going down, but the adjusted score trend is only just going down.

It's a much flatter line.

We see science scores slightly increasing with the adjusted line.

Also slightly increasing, but less quickly.

And we see a flat line for reading scores where they were increasing otherwise.

Scott writes, but these are American scores only.